Unhandled Memcached Error 0x8 When Fetching Data from Couchbase

[mananshukla-inventyv]

Issue Description
We are encountering an unhandled memcached error (0x8 - NO_BUCKET) when attempting to retrieve a document from Couchbase. The issue appears intermittently, and resetting Couchbase connections does not resolve it.

INFO - Reusing server server-node-cb-5.xxxxx:11210 (0x7ff46400b0e0). OldIndex=6. NewIndex=6
ERROR - Got unhandled memcached error 0x8
WARN - server-node-cb-5.xxxxx:11210 (CTX=0x7ff464001480,memcached,SRV=0x7ff46403fc10,IX=2)
Received server error NO_BUCKET (0x8) on packet: OP=0xbb, RC=0x8, SEQ=2
ERROR - Error in getting data from Couchbase:
"A generic / unknown error happened: {"collection":"_default","status":8,
"key":"ACT_TT_6522221a-8037-4a04-97df-fxxxf398b46a1","opaque":2,
"bucket":"cb_bucket","scope":"_default",
"remote":"server-node-cb-5.xxxxx:11210"}"
INFO - Resetting Couchbase connections
INFO - Version=3.0.3, Changeset=0xdeadbeef
INFO - Requested network configuration: heuristic
INFO - Requesting connection to node server-node-cb-5.xxxxx:11210 for CCCP configuration `

[chvck]

Hi @mananshukla-inventyv this looks like you're using the version of the rust SDK in the current master branch, which is built on top of libcouchbase? We are no longer developing this version of the SDK. (https://github.com/couchbaselabs/couchbase-rs?tab=readme-ov-file#deprecation-notice). You'll find a new version of the SDK that is in active development in the nativex branch, we'd love it if you'd be interested in trying it out.

[mananshukla-inventyv]

Hello @chvck ,
Thank you for your response! We’re eagerly awaiting the release of the new native Couchbase SDK version and would love to try it out.

But, we are using Couchbase servers in live applications and are constrained to using the alpha sdk (1.0.0.alpha-4) which was last released on crates.io, as it’s the only one that compiles and works in rust at the moment. Our current Couchbase server version is 7.6.2, and we’d be happy to provide you with the code for reference.

Could you please assist us with the issue we’re facing? Additionally, could you share an estimated timeline for when the new native SDK version will be fully developed?

[chvck]

Hi @mananshukla-inventyv it looks like you've found a way to handle this based on your forum post?

Additionally, could you share an estimated timeline for when the new native SDK version will be fully developed?

Tentatively I think we're targeting GA at mid to end of this year. At this point it will have also gone through our more rigorous set of additional testing (testing behaviour across rebalances etc..) too.

[mananshukla-inventyv]

Hello @chvck,

We have found a workaround for the issue by altogether stopping the cluster connection reset logic and keeping only bucket connection resets when we retry from the application side.

We've found that resetting cluster connections from the application side is an effective way to mitigate timeout errors under heavy loads. We've been successfully applying this approach with the Node.js SDK for quite some time. However, due to the memcached errors caused when the cluster takes time to establish a bucket connection, using long timeouts is not an ideal solution and ultimately needed to remove the cluster reset logic.

If you could address this issue in the current working alpha version of the Rust SDK, it would be a great help until the stable version is released.

Thank you!